IT infrastructure assessment?

Question

linux

Hello!
I would like to evaluate in my small enterprise the reasons for the brakes of 1c applications, subd and vmware. those. comprehensive performance evaluation.
How much will it cost if I turn to knowledgeable offices?
If you do it yourself, what tools are required for this. Maybe someone did something similar?

Answer 1

I am not a certified technology expert, nor have I taken any courses like this I can only speak from my experience. It is quite possible to evaluate and improve the performance of your system without an external contractor.
There are two simple tools:
1) in order to understand whether everything is in order with the iron - Gilev's test. It's a database that tests CPU, RAM, and ROM and compares results to other users on similar hardware. If your data is worse, then this is a clear call that you need to upgrade your equipment or improve your DBMS settings if you have a server base.
2) performance measurement in debug mode. It does not show a lot of what is happening in the system (especially in the client-server model), but it showed itself very well for evaluating thin places. This is a great help to improve your own (non-standard) solutions.
There is also a MCC, but to use it, you need to read the manual supplied with the program and deal with the difficult settings of the technological log. For large enterprises, this is of course a necessary thing as monitoring the entire system and catching problems that are difficult to reproduce. But when optimizing the presence of a specific performance problem, I consider it an overly complicated tool.
To raise morale, I can share three cases from practice.
1) They asked me to speed up filling the document with data - it would seem that there is very little information in the system, but on good hardware, the selection algorithm worked out for several minutes. With the help of performance measurement, I found out that the algorithm is redundant - initially a large table was collected, a small selection was made from it, a large table was destroyed and multiple subqueries were made on the elements of a small selection, which formed a structure similar to the original table and which took all the time. A couple of hours of modification of the algorithm and selection now takes 10 seconds.
2) Once upon a time I was a system developer for a small Ukrainian distributor. A few years later it was already a national company with branches in all areas, and they invited me to solve the exchange problem. They had an RDB that now took all night to collect data from branches. Almost all of my code has remained untouched over the years and I already knew where to optimize - it just didn’t make sense before, since the exchanges fit into an hour or two. I agreed with the project manager and simply threw out unnecessary data from the exchange, raising the speed immediately by 2-3 times.
3) Developed an online monitoring system. The processing of the main dispatcher collected data and displayed it for as long as 15 minutes, which was extremely slow. With the help of performance measurement, I found out that most of the time is spent communicating with the DBMS. then I rewrote the algorithms to use tables with pre-prepared data in RAM - this allowed me to speed up from 15 minutes to 4-5. Further, Gilev's test confirmed with his parrots that the problems were in the gland. We used a virtual machine on Ukrainian hosting. We rented for the same money from Hetzner in Germany and immediately got a double increase in performance - the monitor began to work out in 1-2 minutes. Next, we tweaked the parameters of the virtual machine itself a little (experimented with different types of virtual CPU,

Answer 2

Use a special MCC conf (available on torrents ), it will show how the database and server cluster feel.
The rest below the level and other utilities can also be monitored.

Answer 3

There is nothing to evaluate.
The specialist will come and immediately begin to do (for money, of course).
Anyone who just "for viewing" takes the loot - a scammer.